The A/V Thread

[flowersofnight]

How about a different kind of graph? I tried capturing a 400Hz reference tape with the CR-70 and DR-2 and normalizing both to -6dB for a head-to-head comparison:

CR-70:


DR-2:


Computer-generated pure 400Hz sine wave:


The overall noise level is pretty similar apart from the DR-2's spikes in the high end. The CR-70's high end bends up uniformly a little more instead - maybe those are just two different ways of handling the same phenomenon.
The "side spikes" on the CR-70 are odd. They are at precisely 35Hz above and below the main 400Hz peak, so it doesn't seem random. I wonder if this is a problem or what - is something 54dB below the peak even audible?

 

[flowersofnight]

Counterpoint: at 3000Hz the CR-70 is unmistakably better.

CR-70:


DR-2:

 

[cardy]

I need to get a testing tape(s) to see if my deck is acting funny.
Would you recommend an azimuth specific tape, you shared a link to one last year, or is better to go with a specific tone tape like you have here.

I believe the tape you linked is the one being sold by this person:
https://naks.es/calibration-tapes/

The thing I wonder about....is if all decks / players have some sort of spike in their frequency range and will never be perfect like the computer generate tone shows?

 

[flowersofnight]

Would you recommend an azimuth specific tape, you shared a link to one last year, or is better to go with a specific tone tape like you have here.

What exactly are you troubleshooting?
Azimuth tapes are good for nothing else except calibrating azimuth, they're recorded in a different way from normal tapes. And if you are doing azimuth, you also need a 1kHz head height tape.

If you want the naks.es tapes I'll sell you mine for che$p - slightly used of course. I traded up to different tapes from ANT Audio and GennLab in the process of wrasslin' with this CR70.

The thing I wonder about....is if all decks / players have some sort of spike in their frequency range and will never be perfect like the computer generate tone shows?

I believe the spikes at the first harmonic and second harmonic, with the second being stronger, is something inherent to tape decks and their magnetic heads. Notice that the 400Hz plot has spikes at 800 and 1200Hz, and similarly with 3000/6000/9000. The other things, like the suspicious 15.7kHz spike on my DR-2, I don't know.
Of course the computer tone isn't perfect either due to quantization noise - a real perfect tone would just be a single sharp spike. I'm too lazy to actually do this but I think if you keep cranking up the bitrate on your generated WAV you can keep pushing down the noise. But even the one I posted has noise well below what you'd get from any physical device.

 

[sanctum]

Well butter my butt and call me a biscuit, I had no idea that VHS were still being produced, but Mortiis just went and announced a special repress of his 1997 video release.

 

[flowersofnight]

30 min music video

What, only one song?

I don't know about VHS, but my understanding is that as far as modern-day cassette tapes go, the quality ranges roughly from "abysmal" to "about as good as a decent Type-I tape from back in the day". And new vinyl is trash half the time because they're just wall decorations that they know nobody will ever play. I'm curious if VHS will be any better.
Some of the calibration cassette tapes I have are recorded on this stuff:
https://www.recordingthemasters.com/blank-cassettes

 

[sanctum]

And new vinyl is trash half the time because they're just wall decorations that they know nobody will ever play.

If it is put out by Back in Black then one shouldn't get their hopes up regarding the quality.

I have a couple of Mortiis albums on vinyl and they sound superior to the digital versions that I had been listening to until I switched. Emperor also re-released all of their albums as 'half speed masters' not too long ago and I've heard that those sound incredible.

What, only one song?

His early releases were one big song, split into two parts, with one part occupying each side of the tape/vinyl.

 

[flowersofnight]

Some notes on scanning: Japanese indies magazines used cheap, lousy paper that is see-through.

Scanned in the ordinary way, you see the next page showing through very clearly.
But we know that this is really a cheap publication that only had one color of black ink, so we can just play with the black/white levels and say black is the black of the headline text, and white is the level of the "next page grays".



This improves things but you can still see writing on Klaha's shirt. We can improve the original by putting a black sheet of paper behind it while scanning to minimize "bounceback":



If you do that and then pull the trick with the levels again, maybe that's the best you can do. In principle one could even go back and "re-gray" the image to make it paper-colored again, but I'm not sure if that is silly or what.

I've read that some people will actually take the scan of the opposing page, line it up precisely with the original scan, and subtract it out. But it seems massively impractical to align every pair of pages precisely and get the "subtraction factor" just right for each and every one. Even juggling sheets of black paper is already a pain.

 

[Js_klaus]

Just found this tutorial for video capturing, Im actually trying some of the tips

 

[flowersofnight]

Just found this tutorial for video capturing, Im actually trying some of the tips

What is this nonsense? Capturing 1080p 60fps video for VHS tapes? In composite? Directly to compressed video?

Here's my advice instead. And I'll even write it in text form that you can read in 30 seconds, rather than a video that takes 30 minutes.

Video hardware
For the love of God, at least find something with an S-Video output, not the composite they were using in that Youtube tutorial. (And not the component output like on some DVD+VHS decks either) Some high-end VCRs have built-in timebase correctors (of sorts) which will improve quality - get one if you can. There's a lot of "magic beans" thinking in the VCR space, but just get something decently good that fits your budget, and clean the heads as recommended in the manual.
If you find yourself needing anything more exotic like external TBCs or Macrovision strippers, you'll probably know it.

Capture hardware
The ATI TV Wonder 600 USB is generally recognized as the best, but it's old, hard to find, and won't cooperate with versions of Windows newer than about Win10 build 2004. I think I've read that devices with Magewell chipsets are the best of what's currently out there but I haven't tested this.
Be careful with "gaming" hardware, as some of them (such as Micomsoft's XCAPTURE) don't handle dropped frames properly and just show a blue screen or similar.

OS check
You may want to use LatencyMon to make sure your combination of system/drivers is not behaving weirdly and being prone to drop frames. For example if you use the ATI TV Wonder on a new version of Windows, this check will fail.

Capture software
VirtualDub is the simplest and best for this. Some people swear by the 1.9.x versions over the 1.10.x versions, but I suspect that's voodoo superstition.
You might even be able to get away with just capturing via FFMPEG these days (with the appropriate settings) if you're a command-line jockey.
OBS can capture video (as you saw) but I suspect it needs a lot of wrangling to behave like an archiving application rather than a Twitch streaming application. I don't see the point when there are alternatives more suited to our purpose.

Capture settings
NTSC video is only recorded at 480i at 29.97fps - anything more than this is pointless overkill. I capture at 720x480 which is my capture device's native resolution.
Some capture devices can set the black level differently for either NTSC-U (US) or NTSC-J (Japan). Probably best to match this to your material, though I've never noticed any difference either way.
You should use uncompressed codecs when capturing, so you can do any editing/trimming/postprocessing without losing quality. Then compress only when you're ready to release the final product. HuffYUV is still good for video, and audio you may as well just leave as WAV. UTVideo seems to be a newer alternative to HuffYUV but I haven't tried it - modern systems are fast and big enough that capturing SD 30fps video just isn't very demanding. If you're capturing 8K 120fps footage or something, the choice may be more important. Lagarith seems like it's junk and always has been.

Good luck

 

[Js_klaus]

What is this nonsense? Capturing 1080p 60fps video for VHS tapes? In composite? Directly to compressed video?

Here's my advice instead. And I'll even write it in text form that you can read in 30 seconds, rather than a video that takes 30 minutes.

I didn't watch the whole thing but I'm testing this 29.97 to 59.94 and it's actually helping me a lot with the interlacing artifacts, still more testing needed.

 

[flowersofnight]

but I'm testing this 29.97 to 59.94 and it's actually helping me a lot with the interlacing artifacts

So you're capturing each field to its own frame?

 

[cardy]

What is this nonsense? Capturing 1080p 60fps video for VHS tapes? In composite? Directly to compressed video?

Here's my advice instead. And I'll even write it in text form that you can read in 30 seconds, rather than a video that takes 30 minutes.

Video hardware
For the love of God, at least find something with an S-Video output, not the composite they were using in that Youtube tutorial. (And not the component output like on some DVD+VHS decks either) Some high-end VCRs have built-in timebase correctors (of sorts) which will improve quality - get one if you can. There's a lot of "magic beans" thinking in the VCR space, but just get something decently good that fits your budget, and clean the heads as recommended in the manual.
If you find yourself needing anything more exotic like external TBCs or Macrovision strippers, you'll probably know it.

Capture hardware
The ATI TV Wonder 600 USB is generally recognized as the best, but it's old, hard to find, and won't cooperate with versions of Windows newer than about Win10 build 2004. I think I've read that devices with Magewell chipsets are the best of what's currently out there but I haven't tested this.
Be careful with "gaming" hardware, as some of them (such as Micomsoft's XCAPTURE) don't handle dropped frames properly and just show a blue screen or similar.

OS check
You may want to use LatencyMon to make sure your combination of system/drivers is not behaving weirdly and being prone to drop frames. For example if you use the ATI TV Wonder on a new version of Windows, this check will fail.

Capture software
VirtualDub is the simplest and best for this. Some people swear by the 1.9.x versions over the 1.10.x versions, but I suspect that's voodoo superstition.
You might even be able to get away with just capturing via FFMPEG these days (with the appropriate settings) if you're a command-line jockey.
OBS can capture video (as you saw) but I suspect it needs a lot of wrangling to behave like an archiving application rather than a Twitch streaming application. I don't see the point when there are alternatives more suited to our purpose.

Capture settings
NTSC video is only recorded at 480i at 29.97fps - anything more than this is pointless overkill. I capture at 720x480 which is my capture device's native resolution.
Some capture devices can set the black level differently for either NTSC-U (US) or NTSC-J (Japan). Probably best to match this to your material, though I've never noticed any difference either way.
You should use uncompressed codecs when capturing, so you can do any editing/trimming/postprocessing without losing quality. Then compress only when you're ready to release the final product. HuffYUV is still good for video, and audio you may as well just leave as WAV. UTVideo seems to be a newer alternative to HuffYUV but I haven't tried it - modern systems are fast and big enough that capturing SD 30fps video just isn't very demanding. If you're capturing 8K 120fps footage or something, the choice may be more important. Lagarith seems like it's junk and always has been.

Good luck

Great summary. I agree with everything here and I pretty much follow these steps with slight variations (different USB card, I use Lagarith codec)

The Panasonic ES-10 is a great cheapo device ($25-100 used) which often helps stabilize the signal from a VHS deck. In lieu of a proper TBC like the Datavideo 1000 which are too expensive and too risky to buy these days, that ES-10 has worked wonders.
I actually have the Datavideo 1000 and it couldn't fix some screen tearing issues which the ES-10 magically fixed.

Allowing the capture card to go directly to 59.94fps can result in the bitrate being too low and runs the risk of introducing compression artifacts. And of course capturing to mp4 / h.264, or before that DVD/mpeg2, is a lossy format, like saving scans directly as jpeg files. For bright well-lit content, that might be good enough but for more challenging content, capturing at native sizes and with Huffyuv or Lagarith will allow for more flexibility since you most likely plan on resizing / color correcting / fixing things up later and don't want to have too much recompression of lossy formats going on.

I capture at 720x480i @ 29.97fps / Lagarith codec and eventually up-res to 1440x1080 and 59.94fps export to 10bit h.265.

This helps solve a few issues for creators / consumers of such content:
-Most modern software does not deinterlace well or even recognize deinterlace flags in mp4s. VLC will deinterlace a DVD properly but I've had mixed results playing back interlaced mp4s depending on how to spec you encode. Some programs don't even recognize the interlacing flag and will just play it back at 29.97 progressive and not show both fields. Going to 59.94 makes sure you get that smooth playback that isn't dependent on the software.

-Encoding to 480i interlaced in mp4 is perfectly doable but the presets in many programs just look at the framesize and allocate a measly amount of bitrate. If your original VHS source is very noisy or has a lot of motion, as is the case with many Japanese music tapes with shaky live footage, then some presets in Handbrake or CQ don't seem to give enough data to encode the content. Going to 1080p and using an appropriate bitrate makes sure there's less compression artifacts.

-Future proofing. Encoding to these higher settings makes you less dependent on whether software will playback interlaced video properly. If the videos are intended to be distributed, most people below a certain age won't even know what interlacing is since VHS has been gone for a generation now. It's too legacy to deal with and I'm of the mindset of getting rid of for good. I know this sounds odd since I am a believer of maintain the source intact but interlacing is such a quagmire that presents so many visual problems that in most cases negatively detriments the playback for most people who try viewing it.

-One can argue that h.265 is excessive and is a processor hog but I consider it like the 320kbps mp3 to my FLAC source. It's really darn close to the source at a fraction of the size and with a pretty broad software / hardware compatibility. I could essentially reencode that h.265 file to h.264 if I really get concerned with playback support but until then I can enjoy the visual / compression benefits of h.265.


I recommend using AVIsynth and QTGMC to deinterlace.
The results are far superior to any realtime software based deinterlacer and better than what Premiere Pro / handbrake can do to deinterlace.

While long, this video explains how to use that tool perfectly:

There's a learning curve due to AVISynth being a scripting language but once you get the basic boilerplate down, you can pretty much reuse those same settings for every file.

This Youtuber is an example of doing analog capture right.
https://www.youtube.com/c/ReelyInteresting

He also scales up to 59.94fps and 4K video which is really just to ensure that youtube gives the appropriate bitrate since it starves 480p content.
I'm almost certain he's using QTGMC to deinterlace.
He uses a denoiser which I'm a bit iffy about (grain doesn't detract that much to me but denoising often blurs edges / details) but overall his capture / conversion results are the best I've seen on youtube.

 

[flowersofnight]

Most modern software does not deinterlace well or even recognize deinterlace flags in mp4s

Any idea if the same is true for AVIs? I typically just leave everything in AVI since that's what I capture it in and there doesn't seem to be any particular reason to jump formats.
Lately I've been just letting the chips fall where they may as far as deinterlacing goes, figuring "The player will handle it", but I wasn't familiar with QTGMC. If there are now better deinterlacing algorithms that are too slow for realtime, maybe it's worth revisiting some of those captures I did recently and hard-deinterlacing them. I think I'll go back too and change the VBR MP3 tracks to 320K CBR since VBR is technically a rule-break.

He also scales up to 59.94fps and 4K video which is really just to ensure that youtube gives the appropriate bitrate since it starves 480p content.

Anyone uploading video to Youtube is of course Accursed from the get-go, but this is a good point for those who must.

Datavideo 1000 which are too expensive and too risky to buy these days

Yeah, aren't a lot of the "go-to" TBCs actually starting to get capacitor plague and whatnot?
On the internet there's an unwarranted attitude that hardware more than 5 years old is a ticking time bomb, but it seems like a lot of the old high-end A/V gear from the 80s/90s actually is at the point where people are recapping/reflowing them. Like, it's not enough to just get that one magical VCR from 1981 or whatever, now you need one where every capacitor on the S-Video decoder board has been replaced. (Cue the holy wars about what sort of dielectric is best )

 

[cardy]

Any idea if the same is true for AVIs? I typically just leave everything in AVI since that's what I capture it in and there doesn't seem to be any particular reason to jump formats.
Lately I've been just letting the chips fall where they may as far as deinterlacing goes, figuring "The player will handle it", but I wasn't familiar with QTGMC. If there are now better deinterlacing algorithms that are too slow for realtime, maybe it's worth revisiting some of those captures I did recently and hard-deinterlacing them. I think I'll go back too and change the VBR MP3 tracks to 320K CBR since VBR is technically a rule-break.


Anyone uploading video to Youtube is of course Accursed from the get-go, but this is a good point for those who must.


Yeah, aren't a lot of the "go-to" TBCs actually starting to get capacitor plague and whatnot?
On the internet there's an unwarranted attitude that hardware more than 5 years old is a ticking time bomb, but it seems like a lot of the old high-end A/V gear from the 80s/90s actually is at the point where people are recapping/reflowing them. Like, it's not enough to just get that one magical VCR from 1981 or whatever, now you need one where every capacitor on the S-Video decoder board has been replaced. (Cue the holy wars about what sort of dielectric is best )

VLC does playback AVIs with deinterlacing properly if you activate it.
For whatever reason, all of my captures done with Virtual Dub and saved with Lagarith are Top Field First but the file flags it as Bottom Field First. Whenever I turn on deinterlacing in VLC it attempts to do BFF and there's no way that I've figured out how to force TFF deinterlacing in the settings. It's a rare edge case on my part but it's yet another reason why I just deinterlace for good.

I used to encode audio as 320kbps CBR but now use FLAC since the file size difference is negligible. 1 hour of 320kbps mp3 is 150-200 megs? 1 hour of 16bit flac is double that at most?

With the audio I usually have 2 or 3 versions on the same video:
1) Normalized Raw Audio
2) Normalized Audio with DNR applied and maybe some EQ adjustments to make it 'pop'
3) A version with the CD tracks resync if the VHS is a PV one to get CD quality sound

I don't always do that last step as it's a tedious process but even VHS tapes with Hi-Fi audio don't sound all that great when you compare them to the CD audio. Another issue with muxing FLAC audio is that it forces you to use the MKV container. mp4 doesn't support flac. So there are potential capatibility issues but I somewhat think MKV has been universal for the last 10 years or so?

Overall I would highly recommend learning / using QTGMC. When I A-B compared what Premiere Pro does with it's native deinterlacing versus what QTGMC can do, it was like night and day.

All of those old devices are definitely giving up the ghost and getting a clunker is all too real.
I was lucky in that I bought my TBC-1000 20+ years ago and used it sparingly. So far it's still working and the caps look okay. Same goes for my S-VHS deck I bought around the same time. There are people who can repair the TBC-1000 and for VHS repairs, Lordsmurf from digitalFAQ can do that sort of work. I'd be too afraid to try and do that sort surgery myself.

 

[flowersofnight]

Examining the "head alignment" test using an oscilloscope. The typical calibration guide tells you to just maximize the sum of the two channels (red and blue in this graph) but I think there are pathological cases where a far-off alignment actually has a greater sum than a good alignment. More investigation needed. I wish I hadn't already adjusted the output potentiometers, not that it strictly matters.

Picked up a 10K azimuth tape from Ant-Audio to complement the usual 15K one, to make sure I'm not dialing in on the wrong peaks. I think maybe I was, but the head height needs adjustment first I'm guessing.

 

[sanctum]

This is probably a potato-level question, but do speakers get louder as they 'warm up'?
I usually listen to podcasts on my bluetooth speaker in the morning as I make and eat breakfast. It might just be me, but I swear that no matter what I listen to, it sounds louder 10-minutes or so after I push play.

 

[flowersofnight]

This is probably a potato-level question, but do speakers get louder as they 'warm up'?

Not that I'm aware, apart from my powered speakers supposedly wanting a minute or so to warm up when you first turn them on. For the kind you use on a phone or computer, even that seems unlikely.

 

[sanctum]

That speaker does exhibit some other strange behaviours so it might just be odd as the volume output from the speakers I have connected to my record player is consist.